Stoichiometric inconsistency violates universal constraints: 1. Molecular masses are always positive, and 2. On each side of a reaction the mass is conserved. A single incorrectly defined reaction can lead to stoichiometric inconsistency in the model, and consequently to unconserved metabolites. Similar to insufficient constraints, this may give rise to cycles which either produce mass from nothing or consume mass from the model. Implementation: This test first uses an implementation of the algorithm presented in section 3.1 by Gevorgyan, A., M. G Poolman, and D. A Fell. "Detection of Stoichiometric Inconsistencies in Biomolecular Models." Bioinformatics 24, no. 19 (2008): 2245. doi: 10.1093/bioinformatics/btn425 Should the model be inconsistent, then the list of unconserved metabolites is computed using the algorithm described in section 3.2 of the same publication.

This model contains 616 (80.42%) unconserved metabolites: mal__L_c, ura_c, eig3p_c, din_c, duri_c, ...

This will exclude biomass, exchange and demand reactions as they are unbalanced by definition. It will also fail all reactions where at least one metabolite does not have a formula defined. In steady state, for each metabolite the sum of influx equals the sum of efflux. Hence the net masses of both sides of any model reaction have to be equal. Reactions where at least one metabolite does not have a formula are not considered to be balanced, even though the remaining metabolites participating in the reaction might be. Implementation: For each reaction that isn't a boundary or biomass reaction check if each metabolite has a non-zero elements attribute and if so calculate if the overall element balance of reactants and products is equal to zero.

A total of 59 (6.41%) reactions are mass unbalanced with at least one of the metabolites not having a formula or the overall mass not equal to 0: PFK, FBP, HEX7, GTPCI, FRUtpts, ...

This will exclude biomass, exchange and demand reactions as they are unbalanced by definition. It will also fail all reactions where at least one metabolite does not have a charge defined. In steady state, for each metabolite the sum of influx equals the sum of efflux. Hence the net charges of both sides of any model reaction have to be equal. Reactions where at least one metabolite does not have a charge are not considered to be balanced, even though the remaining metabolites participating in the reaction might be. Implementation: For each reaction that isn't a boundary or biomass reaction check if each metabolite has a non-zero charge attribute and if so calculate if the overall sum of charges of reactants and products is equal to zero.

A total of 66 (7.17%) reactions are charge unbalanced with at least one of the metabolites not having a charge or the overall charge not equal to 0: PFK, FBP, HEX7, GTPCI, FRUtpts, ...

Disconnected metabolites are not part of any reaction in the model. They are most likely left-over from the reconstruction process, but may also point to network and knowledge gaps. Implementation: Check for any metabolites of the cobra.Model object with emtpy reaction attribute.

A total of 9 (1.17%) metabolites are not associated with any reaction of the model: selmethtrna_c, RNA_c, dna_c, amba_c, g6p_B_c, ...

A large fraction of model reactions able to carry unlimited flux under default conditions indicates problems with reaction directionality, missing cofactors, incorrectly defined transport reactions and more. Implementation: Without changing the default constraints run flux variability analysis. From the FVA results identify those reactions that carry flux equal to the model's maximal or minimal flux.

A fraction of 17.75% of the non-blocked reactions (in total 101 reactions) can carry unbounded flux in the default model condition. Unbounded reactions may be involved in thermodynamically infeasible cycles: CAT, CCP, FOCYCTCOR, PTAr, GLUR, ...